# huggingface/open-r1

Fully open reproduction of DeepSeek-R1

Repository: https://github.com/huggingface/open-r1
Canonical: https://ross.abutalabs.com/products/open-r1
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-04-02T14:03:15+00:00

## Health v2 (maintenance only)
Score: 54/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 75, release rhythm 35, longevity 41
- inputs: {"age_days": 586, "days_push": 153, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 26447, forks 2445 (observed 2026-08-28T04:11:46.952644+00:00)

## What it is
Open R1 is Hugging Face's fully open reproduction of the DeepSeek-R1 training pipeline, providing scripts for supervised fine-tuning, GRPO reinforcement learning, and synthetic data generation. It includes curated reasoning datasets and recipes to train reasoning-capable language models from base checkpoints.

## Use cases
- reproduce deepseek-r1 training pipeline
- train a reasoning language model with GRPO
- fine-tune an LLM with SFT on reasoning data
- generate synthetic reasoning traces from a distilled model
- build an open reasoning model from a base model
- evaluate models on math and coding reasoning benchmarks

## When to choose
- you want to replicate or build on DeepSeek-R1's training recipe
- you need open scripts for GRPO reinforcement learning of LLMs
- you want curated distilled reasoning datasets like Mixture-of-Thoughts
- you are training reasoning-capable models on GPU infrastructure

## When to avoid
- you just need to run inference with an existing model
- you need a production LLM serving stack rather than training tooling
- you lack multi-GPU resources for large-scale training
- you need a general-purpose fine-tuning framework with broad model support

## Facets
- artifact type: framework
- maturity: active
- function: llm-training, machine-learning, data-generation, rag
- domain: large-language-models, deep-learning, machine-learning, artificial-intelligence
- platform: python
- tags: deepseek-r1, reinforcement-learning, grpo, sft, reasoning-models, distillation, synthetic-data, huggingface, gpu, linux

## Member repositories
- huggingface/open-r1 (main) score 54

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:46.952644+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T16:55:52.199775+00:00, confidence not recorded.
  - readme: https://github.com/huggingface/open-r1 (fetched 2026-08-28T04:11:46.952644+00:00, sha 196cdae31bbb)
- Data as of 2026-08-30T08:39:29.467469+00:00.